1. Field of the Invention
The invention described herein relate generally to transceivers requiring precision quadrature local oscillator (LO) signals.
2. Background
Many communications standards, for example, 802.11a/b/g for WLAN and DVBS2 for Digital Satellite TV, demand fully integrated transmitter and receiver architectures offering high performance at low cost. Direct conversion receivers and image-reject receivers are examples of such architectures. In these architectures, excellent phase and amplitude balance are critical to prevent quadrature inaccuracy from limiting the overall performance. In particular, inaccuracy in the phase and amplitude balance can result in poor image rejection.
In a receiver, image rejection refers to the receiver's ability to reject signals at its image frequency. In a transmitter, poor image rejection can produce an image signal that falls within the receive band of adjacent channels, which will interfere with devices operating on the adjacent channels. The image and desired inputs both mix with the Local Oscillator (LO) signal and are downconverted to the same frequency. This poses a problem in conventional double-balanced mixers because the two downconverted products interfere with each other, since they exit at the Intermediate Frequency (IF) output port together. Image rejection is thus defined as the ratio of the downconverted image signal power exiting at the IF output port, to that of the desired signal, exiting the same IF output port. For example, if the downconverted image and desired signal levels at are −30 dBm and −10 dBm respectively, then the image rejection is 20 dB. As mentioned, good image rejection requires close amplitude and phase matching.
Conventional double-balanced mixers use filters to block the image from entering the mixer. This prevents the mixer from generating a down-converted image signal. It will be understood, however, that as the IF is reduced, the desired and image signals move closer together in frequency, converging on the LO frequency, which limits the effectiveness of filtering and/or increases the complexity and cost of filtering solutions.
In comparison to conventional double-balanced mixers, image-rejection mixers, for example, achieve image-rejection through phase cancellation, not filtering, so the frequency spacing between the image and desired inputs can be negligible. Conventional image rejection receivers can achieve an image rejection ratio in the range of approximately 20–30 dB, which corresponds to a phase imbalance that can be as high as 10°. The standards referred to above, however, can require a phase accuracy of better than 1°. For example, such standards can require 60 dB of image rejection, which can require a phase imbalance that is as precise as 0.1°.
Phase accuracy in conventional receivers is often limited by the on-chip matching of devices within the quadrature generation circuit. Careful layout can help minimize phase inaccuracy; however, typical performance with good matching is still typically limited to ±3°. This corresponds to an Image Rejection Ratio (IRR) of less than 30 dB, which often results in a corrupted signal constellation and high bit error rate.
Given the mismatch limitations of conventional designs, successful implementation of the architectures referred to above requires additional phase calibration either inside the tuner included in the RF front end or inside the baseband demodulator circuit. Many conventional designs use the later approach to address the phase imbalance issue. The problem with such solutions, however, is that the RF front end and the baseband demodulator circuit are often produced by two different entities. As a result, the designer of the RF front end is faced with a dilemma. The designer can assume that the baseband circuit will include the requisite compensation and not include any compensation in the RF front end. But if the baseband circuit does not include sufficient compensation, then the receiver will not perform adequately. Thus, it can be preferable for the RF front end to include the requisite compensation, because it makes the RF front end independent of the baseband circuit. Alternatively, the system design may be forced to purchase both the baseband circuit and RF front end form the same supplier as part of a chip set solution, in order to ensure that the chip set has adequate compensation, which limits the designers options. Further in an implementation of an RF frequency translation modular based on an RF-Analog Baseband-RF conversion architecture for emerging digital satellite TV broadcasting systems, it is imperative to have an analog method to generate an accurate quadrature LO signals for both RF to analog baseband conversion (receiver portion of the FTM modular) and analog baseband to RF conversion (transmitter portion of the FTM modular).
A few analog-based, front-end solutions have been proposed; however, none have implemented continuous phase error correction with <1° performance. One such solution is based on least-mean squared (LMS) algorithm. The main drawback of such solutions are increased power consumption, increased design complexity, and a one-time only calibration run at startup. Therefore, such designs require additional startup time and phase error performance may drift with temperature variation after power-up.
Another solution provides programmable amplitude and phase. The main drawback of such a solution is that they do not include an on-chip sense circuit or calibration engine. Therefore, such solutions still require support from the baseband circuit.
Still another solution generates N harmonics of the LO frequency using a high frequency delay locked loop. The main drawbacks of such an architecture are 1) it uses a frequency doubler for LO generation, 2) it increases phase noise on the LO, 3) it is not optimized for quadrature accuracy with phase error measured at ±5°. As a result, the performance can actually be worse than that achievable without a phase compensation system.